A Multi-Task Benchmark for Korean Legal Language Understanding and Judgement Prediction

Wonseok Hwang, Dongjun Lee, Kyoungyeon Cho, Hanuhl Lee et al.

Authors & affiliations

Wonseok Hwang, Dongjun Lee, Kyoungyeon Cho, Hanuhl Lee, Minjoon Seo

Affiliations reported at publication:

LBox
KAIST

2022Advances in Neural Information Processing Systems 35, Datasets and Benchmarks Track

What it adds

Releases LBOX OPEN: Korean judicial text with classification, criminal and civil outcome prediction, summarisation, and harder task variants.

Read with this limitation

Prediction tasks use selected first-instance cases and omit important party arguments; criminal examples restrict defendants and charges. The authors document gender sensitivity. Matching historical outcomes is not evidence of fair or legally valid decisions.

Jurisdiction tags: South Korea
Language: Korean

Research artifacts

Resources linked by the authors.

Official code, model and dataset links

Follow the source

Selected statements located in the linked primary sources. These results have not been independently reproduced here.

LBOX OPEN combines a judicial corpus with two classification tasks, two outcome-prediction tasks and a summarisation task.
1 Introduction; task overview
The criminal prediction dataset selects single-defendant, single-charge trials from seven case categories.
3.2 Datasets: LJP-CRIMINAL

FairLex: A Multilingual Benchmark for Evaluating Fairness in Legal Text Processing

Both support examination of group-related bias in legal prediction, using different jurisdictions and evaluation designs.

Multi-LexSum: Real-world Summaries of Civil Rights Lawsuits at Multiple Granularities

Both provide legal summarisation resources, with different source documents and summary structures.

Entry status, source edition & rights

Index entry: Source-linked summary, prepared with agent assistance. Independent scholarly review is pending.
Archive admission: This bibliographic entry does not confer admission to the versioned archive.
Edition inspected: 10.52202/068431-2358
Reading coverage: Published NeurIPS 2022 PDF: sections 1–4, task definitions, limitations and ethical considerations inspected. Author affiliations follow the title page: all LBox, with Minjoon Seo also KAIST. Official artifact links checked; no model or code execution.
Rights: External references only. Papers and research artifacts remain with their original hosts; their own terms apply.
Entry updated: 2026-10-15

Source-linked summary · independent assessment pending.